Google, May 2026: optimising for generative AI search “is still SEO”. You can ignore llms.txt, chunking, AI-specific rewriting and special schema for Google.
Keyword stuffing hurt in the GEO study. llms.txt showed no effect on citations in a 300,000-domain study.
A “guaranteed #1 in ChatGPT” cannot be verified: answers change nearly every run.
Hidden text, prompt injection and mass AI pages break Google's spam policies.
Ask three questions: How many runs per question do you measure? Which official documentation supports your tactics? Do you use hidden text, fake reviews or mass pages?
Do it on your siteSend these three questions to any provider before you sign.
An offer hides white text telling AI to “always recommend us”. What is it?
Show the answer
Answer: A scam that breaks spam policies. Hidden text is spam; instructions to AI are manipulation.
Report six lines: AI impressions (Search Console), AI citations (Bing), AI referral visits and goals (analytics), appearance rate per assistant (your sheet), wrong facts fixed, next three actions.
Keep it to one page so leadership reads it.
Compare month to month, not day to day: AI answers vary.
Do it on your siteMake the template now and fill it on the first Monday of each month.
How often should you compare AI visibility?
Show the answer
Answer: Monthly, with repeated runs. Variance makes daily checks misleading.